兰成渝腐蚀管道失效压力的GA-BP 神经网络组合预测方法

为准确掌握管道失效压力，保证管道安全运行，根据神经网络的非线性和良好的函数逼近特性，提出了基于遗传算法（GA）优化的BP神 经网络组合模型的腐蚀长输管道失效压力预测模型。组合模型将最佳组合阀值与权值隐含在网络的连接中，兼具遗传算法、人工神经网络预测 的优点，并克服了原始数据少对预测精度的影响，同时避免了神经网络容易陷入局部寻优的缺陷，也增强了网络的适应性，改善网络的收敛性 ，在客观地反应腐蚀油气管道失效压力变化趋势方面具有一定的优势。通过实例分析，结果表明:BP神经网络的预测值和Modified B31G计算结 果与真实值误差均较大，而GA-BP的预测值与实际结果的相对误差最大为6.12%，有很好的一致性，为管道的预防性维修提供了理论依据。

Combined forecasting method of GA-BP neural network for failure pressure of Lan-Cheng-Yu corroded pipelines

In order to grasp the failure pressure of pipeline accurately, and ensure the safe operation of pipeline, according to the nonlinear and good function approximating properties of neural network, a prediction model on failure pressure of long distance corroded pipeline was proposed by adopting the combined model of BP neural network based on genetic algorithm (GA) optimization. The combined model implies the best combination threshold and weight in the connection of network. It has the advantages of both genetic algorithm and artificial neural network forecasting, and overcomes the influence of less original data on the accuracy of prediction. At the same time, it can avoid the defect of neural network which is easy to fall into local optimization, and also enhance the adaptability of network and improve the convergence of network. It has certain superiority to reflect the change trend of failure pressure in corroded pipelines objectively. Through case analysis, the results showed that compared with the real value, the errors of forecasting value by BP neural network and calculation result of Modified B31G were both larger, while the largest relative error of forecasting value by GA-BP was 6.12%, which has a good agreement. It provides a theoretical basis for the preventive maintenance of pipeline.